MQC vs Traditional Traffic Shaping

tony.parker · ‎05-17-2007

Hi All,

I am getting conflicting advice on which method to use for my WAN.

I have multiple sites with FR connections into an MPLS cloud. The hub end is 1.5M, with spokes of 768K, 512K, 384K, and 256K.

The requirement is for cisco voice traffic mixed with data.

Can anyone detail the functional differences between using an MQC approach using "shape Average" or "frame-relay traffic-rate" commands within policy maps, then finally applied under the DLCI using the "class" command, versus having a map-class with CIR etc. set and "frame-relay traffic-shaping" applied at the interface and the map-class applied at the sub-interface.

Does the queueing ultimately happen the same way?

Please let me know if more info is needed.

Many Thanks.

Manoj Wadhwa · ‎05-17-2007

Hi Tony,

As your spoke routers have BW of less that 768K, as per Cisco recommendation, you will have to use Fragmentation as you are also supporting voice. Using the FR map-class will give you the flexibility to specify the serialization delay and Tc parameters to suit your voice traffic. As you will also use Traffic shaping, you can configure a LLQ for your voice traffic too. Hence using FR map-class will work out the best solution in this case. The configuration in a nutshell would include traffic shaping, FR fragmentation, specifying serialization delay etc. Hope this helps. Thanks!

- Manoj

tony.parker · ‎05-17-2007

Hi.

Thanks for the reply.

Both scenarios use a map class. My question is whether to apply it to the the sub-interface, or under the DLCI.

Also, as the traffic is going into an MPLS cloud, I have been explicitly told NOT to use fragmentation. Even if I did, as far as I can see you can only apply one fragmentation command to each DLCI, therefore requiring it be set to a size appropriate for the lowest speed link (256K). This will compromise the faster links.